You are referring to AMDs Lightning Bolt / DockPort. That is just a good way to save some space on ultrabooks, because it uses on connector for DP and USB. It does not have PCIe capabilities.
so use 2 USB 3.0 lanes? That should be enough bandwidth to get GPU performance that's superior to a GTX860M at least.
idk, I'd really like to see standardized docking ports for the consumer laptop market that are capable of housing a dedicated desktop GPU. An underside PCI-E lane is probably the best solution, just have to get every manufacturer to agree to one standard, which they've done before.
At the sizes and estimated price of some of these external GPU enclosures, an mITX system based around a silverstone RVZ01 offers quite a bit of competition. and with mITX, the only thing you don't get is a compact monitor
Since when do they ever? For example, MicroUSB 2: No, lightning connector up to iphone 4s, then (whatever it is called) to iphone 6 & 6+. even though it is practically universal.
And yet the motivation from Intel was the high cost of optic components. Their implementation with copper is already too expensive for the mass market and according to Intel themselves, it would have been even more expensive to go for an optical solution.
A simple passive optical cable might be cheaper than the current active copper Thunderbolt cables, but you still need to put a transceiver somewhere (and pay for it). Since you can buy a copper Thunderbolt cable for $29, it's safe to assume that the transceivers used cost less than $5 per end. Feel free to point me to an 8-lane (4-channel) 10.3125 Gbit/s optical transceiver that costs less than $10, is super compact, and only consumes 1 W when active. There is no way that fiber would have made Thunderbolt cheaper, and it would have been an absolute bitch to squeeze one of the proposed optical modules plus a first-gen Thunderbolt controller onto a PCB as small as the one in something like a MacBook Air.
I have 32 Gigabytes of ram...two GeForce GTX 680 video cards...and two terra-bytes of SSD space in my desktop computer. Yet...somehow...a $20 Intel Thunderbolt controller and a 5 dollar Mini-display port is deemed "Too Expensive".
I'm having a hard time wrapping my head around that.
the problem is convincing enough buyers that its worth the extra $20 for system builders to start including it as a standard so that we actually have choices that include the option
I'm pretty sure Hrel is more lamenting the lost opportunity to have Thunderbolt as widely implemented as USB is
basically, its a similar deal we have going to LCD tech, where TN is so widely prevalent because its cheap and good enough and consumers don't seem to know any better and/or care (even if they do). USB3.0 is good enough so screw you Thunderbolt, we won't pony up the little extra it might take to get Thunderbolt widely adopted.
Yeah, I know what you mean. I understand they're not going to stick the interface on a mouse, or even a cheaper monitor, but for certain applications (like this) this shouldn't be a big deal.
That's what they do on every single article TEAMSWITCHER, they play banker and bean counter and whiner. They all think they are privvy and smart enough and in the loop to be in control of entire industries and multi-billion dollar company policies. They read some clueless speculator then babble like a robotic parrot. That's why it never makes sense.
Developers who sign a developer agreement get access to specs, development tools, and reference designs. There is no license fee per se.
Developers must get their products certified before they can use the Thunderbolt logo and sell their products. The certification process does cost to some degree, as testing is performed by independent labs. These labs also can provide test services for design verification and troubleshooting assistance.
As the Thunderbolt controllers are only provided by Intel, there are no alternate sources, but such is the case for many tech components.
The prime barriers to entry are: - development cost - design complexity - small market
Call it what you will, but it is not a licensing fee.
That being said, there are quite a few developers pursuing Thunderbolt. It is a specialty market. See www.thunderbolttechnology.net for a cross section of certified products.
There is a serious disconnect between BOM cost and retail pricing for Thunderbolt devices that is very hard to explain in the absence of some sort of manipulation by Intel. The design complexity and number of external components required for a 2-port device plummeted once Intel integrated a DisplayPort redriver in the Cactus Ridge controllers. According to ARK they only cost $9.95 apiece, but we have yet to see device prices drop below $90 per port (except for the Belkin Thunderbolt Express Dock, which they seem to be fire saling for $159 right now). Even with the other components that are still required for a minimal design (PMICs, DP to Dual-Mode DP converter, power supply, etc.), I don't see why a single slot external PCIe enclosure needs to cost more than $200.
It works with a small modification on OS X driver. I have tested and i use that with OS X since many month.
I have tested with a Thunderbolt->ExpressCard and an ExpressCard to PCIe (limited to PCIe 2.0 x1) and with a TH05, a Thunderbolt to PCIe 1.1 x4 and no problem.
The only problem is that you can't hot swap the card.
Kristian Vättö You should have shown some results with 780TI (which is plug and play)/R9 290x running on a 1x PCI-E 2.0 from express port. The extra hardware (400W PSU + PE4L 2.1b\EXP GDC ) is ~100$ extra hardware, and its maxes out most games. TB2 is if you want to max out the other 5% games which are actually bandwidth heavy.
just dont buy a ridiculously expensive case. Buy a PCIe Riser to Thunderbolt Cable, add in a cheap 400w power supply, and slap it in a cheap mini itx case. Plus you dont have to have a 780 Ti. A 760 should do nicely.
Really hope this comes down in price a bit more. Pretty much turns a laptop with a solid CPU + Thunderbolt 2 connector (13in Retina?) into the ultimate computing situation. Use it on the couch or on the road, and its cool, portable, silent, comfortable. Get home, plug it into the base station and a big monitor, and its a powerful gaming machine too.
Exactly. Cannot comprehend why worrying laptop Manufacturers have not grasped such rather novel usage for an perfectly attainable tech since a while, and tried to sell it already...
It seems easier just to buy a laptop with a secondary gpu already in it? Then you have it when you travel. The price difference is probably less that what you have to pay for this. I have an NVidia 650M in my work laptop, and it games reasonably well, and is with me in my hotel room when I travel.
Kristian you should take a look at the expresscard/mPCI-E results (the majority of results on TechInferno). The adapter (PE4L 2.1b) only costs $70, and Nvidia drivers switch to a compressed PCI-E mode when detecting an x1 link, roughly doubling performance up to thunderbolt levels despite half the bandwidth. Typically synthetics on an x1 2.0 link are slightly faster than thunderbolt, but games are usually slightly slower.
It's still a quite limited bandwidth and so it will cripple most high-end cards. So we're left with an external solution that needs a high-end card to make sense (since integrated graphics is getting better by the day) while at the same time the better the external card the more starved on bandwidth it will be relatively speaking.
No, the most compelling reason for this over integrated graphics is: HEAT.
Most normal laptops can't handle CPU + GPU heat for gaming. Offloading the graphics load to a dedicated card with it's own thermal ecosystem is perfect for laptop gaming. Maybe not in an airport or on an airplane, but at home, or in the hotel, it would be great.
The logistics aren't the issue hereDIYe fact that we have to resort to DIY at all is... Intel should be all over this kinda thing, pushing for it to happen, not like they have a discrete GPU business to cannibalize. If anything this would sell a lot more Intel laptops if it was easier to setup and mainstream.
Some of the performance hit may come from the fact that the ThunderBolt PCIe lanes are usually connected to the chipset and not directly to the CPU the way most GPUs are. Not that this is going to change any time soon.
Actually, just the opposite is true. The lion's share of PC's with Thunderbolt ports are Macs, and most Macs have the Thunderbolt controller connected to PCIe lanes coming from the CPU, not the PCH. I think iMacs with discrete graphics are the only examples of Apple using the PCH lanes instead.
No matter how you slice it, 1.385 GB/s (the real world PCIe throughput limit of Thunderbolt 2) is less than ~13 GB/s (the maximum throughput of a typical PCIe x16 slot). However, PCIe traffic is often bursty, and like memory bandwidth, with decent caching, after a certain point you see seriously diminishing real world performance benefits from further increases in bandwidth alone. Latency is certainly a factor, though, and Thunderbolt is pretty brutal in that regard—around 1.5 µs per hop versus a few hundred ns for a typical PCIe switch. Obviously you would want a Thunderbolt connected GPU to be the first device in a chain, or else you could be facing as much as 9 µs of round-trip latency for the last device in a 6-device chain. If GPU drivers were in any way optimized for lower bandwidth / higher latency connections like Thunderbolt, I'm sure the performance gap could be narrowed even further though, despite having 1/10th the bandwidth and 5x the latency.
This was more of a "hey, I found this video and thought others would like to see it" kind of a post, not a detailed look into eGPUs and all the available options. The ViDock is an alternative option, although the guys over at Tech Inferno doesn't seem to be big fans of it. If I had the tools, time and knowledge I would certainly go deeper and test some of these solutions but unfortunately I currently have none of the three requirements.
OK but your article's title sells it more as an intro to the world of eGPU and as such surely needs to at least mention its existence (enclosure including PSU and expresscard cable $279 for GTX 770 capable model). I've been following eGPU developments for the last couple of years, and ViDock is clearly the sweet spot in terms of cost vs. simplicity to set up, for cards up to GTX 770.
I have no particular connection to Village Instruments, and have not tried their products, but reading Tech Inferno's intro, my interpretation is that they seem obliged to manufacture reasons why their DIY community-led solution is better than ViDock because their products render said DIY solutions (up to GTX 770) obsolete.
Is this a joke? Everyone can see what the title says. I am talking about what the title means. Every sentence ever created is meant for others to read things into it. That is what language is.
Tegeril is correct that you sold yourself on the idea of what this article was about. No, not every sentence is meant for people to read things into it other than exactly what the sentence states. Sometimes people imply something but sometimes a sentence is meant to convey nothing more than exactly what it says.
I don't fault your for thinking there might be something more to it without any context other than the title itself but, apparently, even after you read the article you STILL did not have enough context to determine that your interpretation was incorrect even though the title is a 100% spot on description of what the article is about if you don't try to read some sort of implied meaning in it. Context, learn to use it better or you are going to spend your whole life disappointed.
ViDock is not a certified Thunderbolt developer meaning Intel will give him no Thunderbolt chips.
He only has 5Gbps expresscard enclosures requiring you to purchase a $150 Sonnet Echo Expresscard Pro Thunderbolt-to-expresscard adapter to use. That means you get half the total bandwidth if plug into a Thunderbolt1 port or a quarter if plug into a Thunderbolt2 port.
Worse is his expresscard gear costs 3 times more than the DIY eGPU stuff and costs even more than a US$199 10Gbps Thundertek Thunderbolt1 enclosure yet gives only half the bandwidth!! Why would anybody buy his slow, overpriced gear?
The VAIO Z is different in many ways. First, the eGPU is only compatible with the VAIO Z, whereas this and other Thunderbolt setups work with any Thunderbolt-equipped computer. Secondly, it only has an AMD 6650M, which even by mobile standards is a mediocre GPU. With the Thunderbolt method your GPU choices are much broader and you can go all the way to the fastest desktop cards like the 780 Ti.
Just a wild thought: maybe there is little interest in external GPU chassis because it would cannibalize the highly profitable mobile GPU market. I mean Intel Iris Pro chips are expensive and so are AMDs and nVidias mGPUs. I'm sure there is little interest for them to do cannibalize their margins on that front.
Why would their be any cannibalization? Any external solution is going to be more expensive than a discrete solution that's inside of a laptop. It would have to include a case, psu and the Thunderbolt logic. Those in retrospect are relatively inexpensive but would allow them to charge a premium, even if the card inside of the case is a simple reference design that desktop users could go out and purchase.
First rule of business: make cool stuff. people will buy it. You are Invidia. Someone buys a desktop GPU from you instead of a mobile GPU. The problem here is?
I think we are all looking the wrong way with these expensive TB enclosures.
What if this type of setup was in fact adopted by monitor manufacturers? I mean, a monitor hasn't got to be portable, has already some type of enclosure (duh!) and also has a lot of unused space in the back for a beefy GPU and a PSU. The whole ultrabook concept is aiming far from gaming on the go, but why not have a thin and light notebook that could also game when you are docked. In fact all ultrabooks could get connected to a Thunderbolt GPU-infused monitor and become decent gaming machines.
An R9-270X-class (aka "decent 1080p gaming", according to AT April 2014) is about 200USD. So, would you pay a premium of, say, 300 USD (200USD GPU + 100USD for PSU / manufacturing / profit) over your standard 1080p monitor price to infuse it with such a GPU?
I thought the issue with this was drive support would be hard to do, but...they already work?
Man, we really need to start seeing some reasonably priced external enclosures!
I checked, and I don't think my Alienware M17x-R4 (circa 2012) has Thunderbolt unfortunately... but going forward I would LOVE to be able to have a solid quad core CPU with good cooling and drive a good external video card that I just leave in place!
Seems like enclosures shouldn't really need to cost more than a normal quality full PC enclosure does?
This needs to FINALLY happen, especially if Nvidia's drivers are already working!
Judging by the performance hike we've seen from the Nvidia 980m gpu in laptops doing something like this is pointless and no longer needed. The 980m sli is heads and shoulders beyond what laptop gamers have been getting with a single 980m only being 8% slower than 780m sli. I wish we'd got this sort of performance increase from the desktop variants.
Can you just please review the MSI GS30. It connects with PCIex16 - so tons more bandwidth then this. And its not that expensive. But I want to know if it can run a 295x2.
I currently run a GTX 970 in an Akitio Thunder2 (~$200) with a dedicated PSU powering the card. It's currently being used from a 13" MBP to color time a feature length Indie film in DaVinci Resolve. Tremendously faster than trying to color on the MBP. I set it up as an experiment and the colorist doesn't want to work without it. The footage is on a large Thunderbolt RAID and there are currently no other TB equipped computers available.
"Given that Thunderbolt 2 offers only 20Gbit/s of bandwidth while a PCIe 3.0 x16 slot offers 128Gbit/s, getting 80-90% of the performance is a lot more than expected."
I'm not sure who wrote that part but isn't that a bit of a red herring? The max theoretical bandwidth of PCIe 3.0 x16 has absolutely no bearing on the topic of running a graphics card over Thunderbolt. The only 2 bandwidth numbers that matter are the max bandwidth of the interface being used (Thunderbolt 2) and the max possible bandwidth the attached device (GTX 780 ti) may try to use. The theoretical max bandwidth of the PCIe 3.0 x16 interface is already absurd overkill for ANY currently existing single card GPU configurations so it gives you absolutely zero baseline to determine how well something would work on an interface of another type.
I haven't seen any tests confirming this, but wouldn't you think that a GPU with more VRAM would reduce the bandwidth requirement for the interface? I'd think that this would allow the game to keep more in RAM, and have less traveling over the PCIe connection. I could be wrong, but this would make sense to me (and it probably depends on how good the game's devs are).
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
Facebook
Twitter
RSS
71 Comments
Back to Article
Hrel - Sunday, May 4, 2014 - link
It's too bad Intel has licensing fee's that make Thunderbolt not just rare to find, but prohibitively expensive to implement.willis936 - Sunday, May 4, 2014 - link
Any low latency 20 Gbps port is going to be prohibitively expensive. Doing it over copper is not trivial.Hrel - Sunday, May 4, 2014 - link
Didn't I see an open source one from AMD during Computex or something? I'm pretty anandtech even posted about it.Death666Angel - Sunday, May 4, 2014 - link
You are referring to AMDs Lightning Bolt / DockPort. That is just a good way to save some space on ultrabooks, because it uses on connector for DP and USB. It does not have PCIe capabilities.Hrel - Monday, May 5, 2014 - link
so use 2 USB 3.0 lanes? That should be enough bandwidth to get GPU performance that's superior to a GTX860M at least.idk, I'd really like to see standardized docking ports for the consumer laptop market that are capable of housing a dedicated desktop GPU. An underside PCI-E lane is probably the best solution, just have to get every manufacturer to agree to one standard, which they've done before.
Zoomer - Monday, May 5, 2014 - link
Latency, CPU overhead. Probably DOA.meacupla - Tuesday, May 6, 2014 - link
At the sizes and estimated price of some of these external GPU enclosures, an mITX system based around a silverstone RVZ01 offers quite a bit of competition.and with mITX, the only thing you don't get is a compact monitor
vortexmak - Tuesday, May 6, 2014 - link
This. An mITX system would come in much cheaper and just a little bit biggerxdrol - Monday, July 7, 2014 - link
You can use a PCIe ribbon cable for that..zacsaturday - Tuesday, April 28, 2015 - link
Apple, agree with something!???Since when do they ever? For example, MicroUSB 2: No, lightning connector up to iphone 4s, then (whatever it is called) to iphone 6 & 6+. even though it is practically universal.
Morawka - Sunday, May 4, 2014 - link
that was intel's choice, it was originally supposed to be over fibre, which is cheaper to makeZotamedu - Monday, May 5, 2014 - link
And yet the motivation from Intel was the high cost of optic components. Their implementation with copper is already too expensive for the mass market and according to Intel themselves, it would have been even more expensive to go for an optical solution.Torashin - Monday, May 5, 2014 - link
I believe there is an optical version available (which can be longer than the copper one).Morawka - Tuesday, May 6, 2014 - link
it would have been faster and by now we would be matching PCIe 3.0 speeds, Driving demand whilst further decreasing manufacturing cost.copper is holding them back to these 10Gbps every 2 years, updates..
willis936 - Monday, May 5, 2014 - link
If 20Gb fiber is prohibitively expensive for consumers then fiber at anything over a Mbps lacks a good adjective for how expensive it is.repoman27 - Tuesday, May 6, 2014 - link
A simple passive optical cable might be cheaper than the current active copper Thunderbolt cables, but you still need to put a transceiver somewhere (and pay for it). Since you can buy a copper Thunderbolt cable for $29, it's safe to assume that the transceivers used cost less than $5 per end. Feel free to point me to an 8-lane (4-channel) 10.3125 Gbit/s optical transceiver that costs less than $10, is super compact, and only consumes 1 W when active. There is no way that fiber would have made Thunderbolt cheaper, and it would have been an absolute bitch to squeeze one of the proposed optical modules plus a first-gen Thunderbolt controller onto a PCB as small as the one in something like a MacBook Air.TEAMSWITCHER - Sunday, May 4, 2014 - link
I have 32 Gigabytes of ram...two GeForce GTX 680 video cards...and two terra-bytes of SSD space in my desktop computer. Yet...somehow...a $20 Intel Thunderbolt controller and a 5 dollar Mini-display port is deemed "Too Expensive".I'm having a hard time wrapping my head around that.
blanarahul - Monday, May 5, 2014 - link
Stay in the closed boundaries of your home, and that is what happens.bunnyfubbles - Monday, May 5, 2014 - link
the problem is convincing enough buyers that its worth the extra $20 for system builders to start including it as a standard so that we actually have choices that include the optionI'm pretty sure Hrel is more lamenting the lost opportunity to have Thunderbolt as widely implemented as USB is
basically, its a similar deal we have going to LCD tech, where TN is so widely prevalent because its cheap and good enough and consumers don't seem to know any better and/or care (even if they do). USB3.0 is good enough so screw you Thunderbolt, we won't pony up the little extra it might take to get Thunderbolt widely adopted.
Wolfpup - Monday, May 5, 2014 - link
Yeah, I know what you mean. I understand they're not going to stick the interface on a mouse, or even a cheaper monitor, but for certain applications (like this) this shouldn't be a big deal.FlushedBubblyJock - Wednesday, March 25, 2015 - link
That's what they do on every single article TEAMSWITCHER, they play banker and bean counter and whiner.They all think they are privvy and smart enough and in the loop to be in control of entire industries and multi-billion dollar company policies.
They read some clueless speculator then babble like a robotic parrot.
That's why it never makes sense.
g4cube - Sunday, May 4, 2014 - link
License, yes. Fee, no.Developers who sign a developer agreement get access to specs, development tools, and reference designs. There is no license fee per se.
Developers must get their products certified before they can use the Thunderbolt logo and sell their products. The certification process does cost to some degree, as testing is performed by independent labs. These labs also can provide test services for design verification and troubleshooting assistance.
As the Thunderbolt controllers are only provided by Intel, there are no alternate sources, but such is the case for many tech components.
The prime barriers to entry are:
- development cost
- design complexity
- small market
Call it what you will, but it is not a licensing fee.
That being said, there are quite a few developers pursuing Thunderbolt. It is a specialty market.
See www.thunderbolttechnology.net for a cross section of certified products.
repoman27 - Monday, May 5, 2014 - link
There is a serious disconnect between BOM cost and retail pricing for Thunderbolt devices that is very hard to explain in the absence of some sort of manipulation by Intel. The design complexity and number of external components required for a 2-port device plummeted once Intel integrated a DisplayPort redriver in the Cactus Ridge controllers. According to ARK they only cost $9.95 apiece, but we have yet to see device prices drop below $90 per port (except for the Belkin Thunderbolt Express Dock, which they seem to be fire saling for $159 right now). Even with the other components that are still required for a minimal design (PMICs, DP to Dual-Mode DP converter, power supply, etc.), I don't see why a single slot external PCIe enclosure needs to cost more than $200.g4cube - Monday, May 5, 2014 - link
Fair argument. Volumes are lower, requiring higher margins to recoup development and marketing expenses.In contrast USB margins range from dollars to just a few pennies.
Wolfpup - Monday, May 5, 2014 - link
Agreed...the prices don't look THAT bad, yet you see basic Thunderbolt hard drives that are hundreds and hundreds over the USB equivalent.Dandu - Sunday, May 4, 2014 - link
FIY, it's possible to make that with OS X : http://www.journaldulapin.com/2013/08/24/a-thunder...It works with a small modification on OS X driver. I have tested and i use that with OS X since many month.
I have tested with a Thunderbolt->ExpressCard and an ExpressCard to PCIe (limited to PCIe 2.0 x1) and with a TH05, a Thunderbolt to PCIe 1.1 x4 and no problem.
The only problem is that you can't hot swap the card.
Dandu - Sunday, May 4, 2014 - link
And in french, with a Thunderbolt to PCIe x4 : http://www.journaldulapin.com/2013/09/23/carte-gra... and http://www.journaldulapin.com/2013/10/13/macbook-a...I play Bioshock Infinite on a MacBook Air at 1440p under Mac OS X with that (and a good graphic card)
egpuFanatic - Sunday, May 4, 2014 - link
Kristian VättöYou should have shown some results with 780TI (which is plug and play)/R9 290x running on a 1x PCI-E 2.0 from express port. The extra hardware (400W PSU + PE4L 2.1b\EXP GDC ) is ~100$ extra hardware, and its maxes out most games.
TB2 is if you want to max out the other 5% games which are actually bandwidth heavy.
schizoide - Sunday, May 4, 2014 - link
Great article. Like many people, I've been waiting for this to actually work and hit a reasonable price point for several years now.Just give me a box with a pcie slot and sell it for $300. It seems like it should be possible! I want a macbook air high-end gaming box!
Morawka - Sunday, May 4, 2014 - link
just dont buy a ridiculously expensive case. Buy a PCIe Riser to Thunderbolt Cable, add in a cheap 400w power supply, and slap it in a cheap mini itx case. Plus you dont have to have a 780 Ti. A 760 should do nicely.Cygni - Sunday, May 4, 2014 - link
Really hope this comes down in price a bit more. Pretty much turns a laptop with a solid CPU + Thunderbolt 2 connector (13in Retina?) into the ultimate computing situation. Use it on the couch or on the road, and its cool, portable, silent, comfortable. Get home, plug it into the base station and a big monitor, and its a powerful gaming machine too.Very cool.
Nexing - Monday, September 22, 2014 - link
Exactly.Cannot comprehend why worrying laptop Manufacturers have not grasped such rather novel usage for an perfectly attainable tech since a while, and tried to sell it already...
kmmatney - Friday, March 6, 2015 - link
It seems easier just to buy a laptop with a secondary gpu already in it? Then you have it when you travel. The price difference is probably less that what you have to pay for this. I have an NVidia 650M in my work laptop, and it games reasonably well, and is with me in my hotel room when I travel.Khenglish - Monday, May 5, 2014 - link
Kristian you should take a look at the expresscard/mPCI-E results (the majority of results on TechInferno). The adapter (PE4L 2.1b) only costs $70, and Nvidia drivers switch to a compressed PCI-E mode when detecting an x1 link, roughly doubling performance up to thunderbolt levels despite half the bandwidth. Typically synthetics on an x1 2.0 link are slightly faster than thunderbolt, but games are usually slightly slower.Calista - Friday, May 9, 2014 - link
It's still a quite limited bandwidth and so it will cripple most high-end cards. So we're left with an external solution that needs a high-end card to make sense (since integrated graphics is getting better by the day) while at the same time the better the external card the more starved on bandwidth it will be relatively speaking.Tegeril - Monday, May 12, 2014 - link
There is a colossally large chasm between integrated graphics and high end cards right now. $150-200 cards run circles around integrated.dflippo - Monday, May 4, 2015 - link
No, the most compelling reason for this over integrated graphics is: HEAT.Most normal laptops can't handle CPU + GPU heat for gaming. Offloading the graphics load to a dedicated card with it's own thermal ecosystem is perfect for laptop gaming. Maybe not in an airport or on an airplane, but at home, or in the hotel, it would be great.
Impulses - Monday, May 5, 2014 - link
The logistics aren't the issue hereDIYe fact that we have to resort to DIY at all is... Intel should be all over this kinda thing, pushing for it to happen, not like they have a discrete GPU business to cannibalize. If anything this would sell a lot more Intel laptops if it was easier to setup and mainstream.SirKnobsworth - Monday, May 5, 2014 - link
Some of the performance hit may come from the fact that the ThunderBolt PCIe lanes are usually connected to the chipset and not directly to the CPU the way most GPUs are. Not that this is going to change any time soon.repoman27 - Tuesday, May 6, 2014 - link
Actually, just the opposite is true. The lion's share of PC's with Thunderbolt ports are Macs, and most Macs have the Thunderbolt controller connected to PCIe lanes coming from the CPU, not the PCH. I think iMacs with discrete graphics are the only examples of Apple using the PCH lanes instead.No matter how you slice it, 1.385 GB/s (the real world PCIe throughput limit of Thunderbolt 2) is less than ~13 GB/s (the maximum throughput of a typical PCIe x16 slot). However, PCIe traffic is often bursty, and like memory bandwidth, with decent caching, after a certain point you see seriously diminishing real world performance benefits from further increases in bandwidth alone. Latency is certainly a factor, though, and Thunderbolt is pretty brutal in that regard—around 1.5 µs per hop versus a few hundred ns for a typical PCIe switch. Obviously you would want a Thunderbolt connected GPU to be the first device in a chain, or else you could be facing as much as 9 µs of round-trip latency for the last device in a 6-device chain. If GPU drivers were in any way optimized for lower bandwidth / higher latency connections like Thunderbolt, I'm sure the performance gap could be narrowed even further though, despite having 1/10th the bandwidth and 5x the latency.
gw74 - Monday, May 5, 2014 - link
Why no mention of Village Instruments Vidock? Their Expresscard eGPU enclosures can be converted to Thunderbolt 2.Kristian Vättö - Monday, May 5, 2014 - link
This was more of a "hey, I found this video and thought others would like to see it" kind of a post, not a detailed look into eGPUs and all the available options. The ViDock is an alternative option, although the guys over at Tech Inferno doesn't seem to be big fans of it. If I had the tools, time and knowledge I would certainly go deeper and test some of these solutions but unfortunately I currently have none of the three requirements.gw74 - Tuesday, May 6, 2014 - link
OK but your article's title sells it more as an intro to the world of eGPU and as such surely needs to at least mention its existence (enclosure including PSU and expresscard cable $279 for GTX 770 capable model). I've been following eGPU developments for the last couple of years, and ViDock is clearly the sweet spot in terms of cost vs. simplicity to set up, for cards up to GTX 770.I have no particular connection to Village Instruments, and have not tried their products, but reading Tech Inferno's intro, my interpretation is that they seem obliged to manufacture reasons why their DIY community-led solution is better than ViDock because their products render said DIY solutions (up to GTX 770) obsolete.
Tegeril - Monday, May 12, 2014 - link
The title says: "Running an NVIDIA GTX 780 Ti Over Thunderbolt 2."You read everything else into that. What you said is not what the title says at all.
gw74 - Tuesday, June 10, 2014 - link
Is this a joke? Everyone can see what the title says. I am talking about what the title means. Every sentence ever created is meant for others to read things into it. That is what language is.masouth - Tuesday, April 14, 2015 - link
Tegeril is correct that you sold yourself on the idea of what this article was about. No, not every sentence is meant for people to read things into it other than exactly what the sentence states. Sometimes people imply something but sometimes a sentence is meant to convey nothing more than exactly what it says.I don't fault your for thinking there might be something more to it without any context other than the title itself but, apparently, even after you read the article you STILL did not have enough context to determine that your interpretation was incorrect even though the title is a 100% spot on description of what the article is about if you don't try to read some sort of implied meaning in it. Context, learn to use it better or you are going to spend your whole life disappointed.
chiaper - Tuesday, May 6, 2014 - link
ViDock is not a certified Thunderbolt developer meaning Intel will give him no Thunderbolt chips.He only has 5Gbps expresscard enclosures requiring you to purchase a $150 Sonnet Echo Expresscard Pro Thunderbolt-to-expresscard adapter to use. That means you get half the total bandwidth if plug into a Thunderbolt1 port or a quarter if plug into a Thunderbolt2 port.
Worse is his expresscard gear costs 3 times more than the DIY eGPU stuff and costs even more than a US$199 10Gbps Thundertek Thunderbolt1 enclosure yet gives only half the bandwidth!! Why would anybody buy his slow, overpriced gear?
gw74 - Tuesday, May 6, 2014 - link
Do thundertek do an egpu solution? what % of performance would a GTX 770 lose through 5Gbps?irev210 - Monday, May 5, 2014 - link
The Sony Z that anadtech reviewed already did just this. Not sure why this is a pipeline story.We need cheaper/more economic solutions for external graphics cards over TB.
Kristian Vättö - Monday, May 5, 2014 - link
The VAIO Z is different in many ways. First, the eGPU is only compatible with the VAIO Z, whereas this and other Thunderbolt setups work with any Thunderbolt-equipped computer. Secondly, it only has an AMD 6650M, which even by mobile standards is a mediocre GPU. With the Thunderbolt method your GPU choices are much broader and you can go all the way to the fastest desktop cards like the 780 Ti.gw74 - Tuesday, May 6, 2014 - link
https://www.facebook.com/VillageViDockGalatian - Monday, May 5, 2014 - link
Just a wild thought: maybe there is little interest in external GPU chassis because it would cannibalize the highly profitable mobile GPU market. I mean Intel Iris Pro chips are expensive and so are AMDs and nVidias mGPUs. I'm sure there is little interest for them to do cannibalize their margins on that front.Kevin G - Monday, May 5, 2014 - link
Why would their be any cannibalization? Any external solution is going to be more expensive than a discrete solution that's inside of a laptop. It would have to include a case, psu and the Thunderbolt logic. Those in retrospect are relatively inexpensive but would allow them to charge a premium, even if the card inside of the case is a simple reference design that desktop users could go out and purchase.gw74 - Wednesday, May 7, 2014 - link
First rule of business: make cool stuff. people will buy it. You are Invidia. Someone buys a desktop GPU from you instead of a mobile GPU. The problem here is?dj_aris - Monday, May 5, 2014 - link
I think we are all looking the wrong way with these expensive TB enclosures.What if this type of setup was in fact adopted by monitor manufacturers? I mean, a monitor hasn't got to be portable, has already some type of enclosure (duh!) and also has a lot of unused space in the back for a beefy GPU and a PSU. The whole ultrabook concept is aiming far from gaming on the go, but why not have a thin and light notebook that could also game when you are docked. In fact all ultrabooks could get connected to a Thunderbolt GPU-infused monitor and become decent gaming machines.
An R9-270X-class (aka "decent 1080p gaming", according to AT April 2014) is about 200USD. So, would you pay a premium of, say, 300 USD (200USD GPU + 100USD for PSU / manufacturing / profit) over your standard 1080p monitor price to infuse it with such a GPU?
Calista - Friday, May 9, 2014 - link
No, I would not since a monitor is often a 10 year investment. A GPU is at most a 3 year investment for any serious gamer.Wolfpup - Monday, May 5, 2014 - link
Wait, so Nvidia's drivers just WORK?I thought the issue with this was drive support would be hard to do, but...they already work?
Man, we really need to start seeing some reasonably priced external enclosures!
I checked, and I don't think my Alienware M17x-R4 (circa 2012) has Thunderbolt unfortunately... but going forward I would LOVE to be able to have a solid quad core CPU with good cooling and drive a good external video card that I just leave in place!
Seems like enclosures shouldn't really need to cost more than a normal quality full PC enclosure does?
This needs to FINALLY happen, especially if Nvidia's drivers are already working!
francescop1 - Tuesday, July 15, 2014 - link
would it be possible to connect the gpu through BOTH the TB ports on a macbook pro for double the bandwidth?henrykale - Wednesday, August 6, 2014 - link
I started testing modulates this week by uploading a few video testimonials and tagging them, now I see why there is so much buzz about the company.nashathedog - Sunday, September 7, 2014 - link
Judging by the performance hike we've seen from the Nvidia 980m gpu in laptops doing something like this is pointless and no longer needed. The 980m sli is heads and shoulders beyond what laptop gamers have been getting with a single 980m only being 8% slower than 780m sli. I wish we'd got this sort of performance increase from the desktop variants.FITCamaro - Wednesday, December 17, 2014 - link
I don't understand why we can't standardize an external PCI-E connector that gives full speed PCI-E to an external device.douglord - Thursday, February 26, 2015 - link
Can you just please review the MSI GS30. It connects with PCIex16 - so tons more bandwidth then this. And its not that expensive. But I want to know if it can run a 295x2.SurfStudio - Tuesday, March 10, 2015 - link
I currently run a GTX 970 in an Akitio Thunder2 (~$200) with a dedicated PSU powering the card. It's currently being used from a 13" MBP to color time a feature length Indie film in DaVinci Resolve. Tremendously faster than trying to color on the MBP. I set it up as an experiment and the colorist doesn't want to work without it. The footage is on a large Thunderbolt RAID and there are currently no other TB equipped computers available.masouth - Tuesday, April 14, 2015 - link
"Given that Thunderbolt 2 offers only 20Gbit/s of bandwidth while a PCIe 3.0 x16 slot offers 128Gbit/s, getting 80-90% of the performance is a lot more than expected."I'm not sure who wrote that part but isn't that a bit of a red herring? The max theoretical bandwidth of PCIe 3.0 x16 has absolutely no bearing on the topic of running a graphics card over Thunderbolt. The only 2 bandwidth numbers that matter are the max bandwidth of the interface being used (Thunderbolt 2) and the max possible bandwidth the attached device (GTX 780 ti) may try to use. The theoretical max bandwidth of the PCIe 3.0 x16 interface is already absurd overkill for ANY currently existing single card GPU configurations so it gives you absolutely zero baseline to determine how well something would work on an interface of another type.
Res1233 - Sunday, August 16, 2015 - link
I haven't seen any tests confirming this, but wouldn't you think that a GPU with more VRAM would reduce the bandwidth requirement for the interface? I'd think that this would allow the game to keep more in RAM, and have less traveling over the PCIe connection. I could be wrong, but this would make sense to me (and it probably depends on how good the game's devs are).